Robust sensing of contact information for detection of the physical properties of an object

Physical properties of an object, such as compliance, are important pieces of information for dexterous handling of the robot hand, and for the locomotion of the legged robot. It is effective to adopt the active sensing method in which a robot pushes the object and senses the reaction force and sequential deformation such as passive rotation of the joint due to the force, to determine the properties. However, the sensors attached to the contact point and the joint tend to break because of an iterative contact or an excessive force. Herein, we propose a method for estimating the force and the consequent joint displacement without attaching sensors to the contact point and the joint. The elastic pneumatic muscle is adopted as an actuator, and the force and joint angle are measured using the inner pressure of the muscle before and after the muscle is extended because of the force. Using the nonlinear model of a one degree of freedom (1-DoF) joint mechanism driven by antagonistic pneumatic muscles, we can estimate the force and the angle, and using the joint mechanism, we can detect in the difference of the compliance of the object. This approach provides the robust robot hand, which can estimate the contact information, including the force and joint displacement, without the need to attach sensors to the fingertip and the joint for example.